3

The hypervolume of the expanded $n$-simplex with side $\sqrt{2}$ appears to be $$\displaystyle\frac{\sqrt{\;n+1\;}\;(2n)!}{n!^3}$$ Does anyone know of a published reference to this result? Or can anyone prove it?


An expanded $n$-simplex can be defined as the vertex figure of an $n$-simplex honeycomb (see the Wikipedia article "simplectic honeycomb") and its coordinates in $(n+1)$-dimensional space can be expressed as all coordinate permutations of $(1,-1,0,0,0,...)$. Low-order examples are the regular hexagon ($n=2$), the cuboctahedron ($n=3$) and the runcinated $5$-cell ($n=4$).

If this is a known result I’d like to cite a reference.

MartinG
  • 978
  • What in the world is "the expanded $n$-simplex"? – Ted Shifrin Aug 14 '14 at 14:48
  • I have added the definition that was given as a comment to my (now-deleted-as-irrelevant) answer. I also TeX-ified the volume expression; please double-check that it's the formula you intended. (I wasn't entirely sure, for instance, just how much stuff was supposed to go under the radical.) – Blue Aug 15 '14 at 16:39
  • Thanks for formatting the expression, Blue. That’s the correct form. Expressed as a multiple of the volume of the simplex with the same side length the volume is (2n)!/n!^2

    = 2, 6, 20, 70, 216,... which is twice OEIS series A001700.

    – MartinG Aug 19 '14 at 19:43
  • That should be 2, 6, 20, 70, 252, ...

    OEIS A001700

    – MartinG Aug 19 '14 at 19:52
  • Some more interesting properties of the expanded n-simplex: – MartinG Aug 19 '14 at 20:01
  • It’s the maximal cross section of the (n+1)-cross-polytope. – MartinG Aug 19 '14 at 20:04
  • It’s also the maximal cross section, and an orthographic projection, of the expanded (n+1)-simplex. This means that by cutting an expanded simplex it’s possible to reveal all lower-order expanded simplices. – MartinG Aug 19 '14 at 20:05
  • The appropriate orthographic projection of the expanded (n+1)-simplex not only delineates the expanded n-simplex but also illustrates, in the projections of its facets, the process by which this simplex is formed by expanding the n-simplex. – MartinG Aug 19 '14 at 20:06

2 Answers2

1

The formula can be derived by decomposing an expanded $n$-symplex in a disjoint union of $*$-rectified symplexes or $*$-rectified hypercubes. Computing the hypervolume of a symplex/hypercube is an easy task, so it is also computing the hypervolume of a $*$-rectified symplex/hypercube.

Rectification is the process of truncating a polytope by marking the midpoints of all its edges, and cutting off its vertices at those points.

Jack D'Aurizio
  • 353,855
  • Would you mind explaining a little more - in particular the terms 'disjoint union' and '*-rectified symplex'? You might use the expanded 3-simplex (cuboctahedron) as an example. – MartinG Sep 03 '14 at 19:51
  • I'm guessing '*-rectified' means rectified, birectified, etc. (and perhaps also '0-rectified'). – MartinG Sep 03 '14 at 20:02
  • 1
    Yes, $n$-rectified means rectified $n$ times, $$-rectified means rectified for a certain number of times (even zero). For "disjoint union" I mean that is possible to find a partition of an expanded n-simplex into $$-rectified symplexes. – Jack D'Aurizio Sep 03 '14 at 20:13
  • The cuboctahedron, for example, can be dissected into $3$ octahedra (rectified cube) and $8$ tetrahedra with side-length $\sqrt{2}$. – Jack D'Aurizio Sep 03 '14 at 20:16
  • Yes, I see that (except that I would describe it as 6 half-octahedra and 8 tetrahedra). And this principle extends to n dimensions in that the expanded n-simplex can be carved out of the n-simplex honeycomb, which consists of -rectified simplices. But I'd be grateful for a bit more guidance as to how to enumerate the various types of -rectified simplex. – MartinG Sep 03 '14 at 20:20
  • 1
    Just have a look at the "Facets per vertex figure" column in the main table of the Wikipedia article you are referring to. – Jack D'Aurizio Sep 03 '14 at 20:23
  • Sometimes, the things we have under our nose are the most difficult to find :D – Jack D'Aurizio Sep 03 '14 at 20:26
  • 1
    Yes, I'd just got there before you pointed it out, and was puzzling over '4 + 4 tetrahedron, 6 rectified tetrahedra' (I think the latter should be halved). But those binomial coefficients are familiar from another approach I've taken to this. If I can show that the volumes of these polytopes, expressed in units of the simplex volume, are another set of binomial coefficients I'll be there. Thanks for pointing out the missing piece. I feel sure someone's done this before, but can't find a reference. – MartinG Sep 03 '14 at 20:41
  • I don't think that table is giving us quite what we need (as illustrated by my note above). The things described by the column headed 'Facets per vertex figure' are not facets but (I think) *-rectified simplexes which the facets cut through (or in some cases enclose). The things we need information on are the hyper-pyramids which connect the facets to the centre, and I don't think the table gives us information on those. – MartinG Sep 03 '14 at 22:19
1

I now have a satisfactory proof of this result of which the following is a summary. The bounty has now expired but I'm still interested to know if there's a published proof.

All the polytopes described below have edge length $\sqrt{2}$.

The n-simplex has coordinates which can be expressed in $(n+1)$-space as the permutations of $(1,0,0,...)$. It has$$ N_{mn} =\binom{n+1}{m+1} $$ m-faces $(m=0,1,2,...n-1)$, where an m=face is a vertex $(m=0)$, and edge $(m=1)$, a face $(m=2)$ and so on. Its volume is $$ v_n = \frac{\sqrt{n+1}}{n!} $$

The expanded $n$-simplex has vertices $\textbf{r}$ which can be expressed in $(n+1)$-space as the permutations of $(1,-1,0,...)$. It can be constructed from the expanded $(n-1)$-simplex by adding two $(n-1)$-simplices whose coordinates are the permutations of $(1,0,0,...)$ and $(-1,0,0,...)$, then shifting these simplices respectively by $-1$ and $+1$ units in the $(n+1)$'th dimension.

It can also be formed by expanding the $n$-simplex, a process which creates an $(n-1)$-face for each $m$-face of the $n$-simplex.

The vertices of the facet so created are those which maximise the dot product of $\textbf{r}$ with the position vector of the centroid of the $m$-face. There are $(m+1)(n-m)$ such vertices, whose coordinates are formed from the ways in which the value $-1$ can be assigned to $m+1$ positions in the vector $\textbf{r}$, and the value $1$ to the remaining $n-m$ positions (or vice versa depending on the relative magnitudes of $m+1$ and $n-m$).

The facet is thus the Cartesian product of an $m$-simplex and an $(n-m-1)$-simplex, and its $(n-1)$-volume is therefore $$ A_{mn} = v_{m}\,v_{n-m-1} = \frac{\sqrt{m+1}}{m!} \frac{\sqrt{n-m}}{(n-m-1)!} $$

Its distance from the centre of the expanded n-simplex is $$ D_{mn} = \sqrt{\frac{n+1}{(m+1)(n-m)}} $$

and its contribution to the volume of the expanded n-simplex is the volume of a pyramid having the facet as its base and height $D_{mn}$: $$ V_{mn} = \frac{1}{n} A_{mn} D_{mn} = \frac{1}{n} \frac{\sqrt{m+1}}{m!} \frac{\sqrt{n-m}}{(n-m-1)!} \sqrt{\frac{n+1}{(m+1)(n-m)}} $$ which can be expressed simply as an integer multiple of the $n$-simplex volume: $$ V_{mn} = \binom{n-1}{m} v_n $$

The total volume of the expanded $n$-simplex can now be obtained by summing over the facets: $$ V_n = \sum_{m=0}^{n-1} N_{mn} V_{mn} = \sum_{m=0}^{n-1} \binom{n+1}{m+1} \binom{n-1}{m} v_n \\ = \binom{2n}{n} v_n = \frac{(2n)!}{(n!)^2} v_n = \sqrt{n+1}\frac{(2n)!}{(n!)^3} $$

By growing the halves of the expanded $n$-simplex outwards from the two sides of an expanded $(n-1)$-simplex it can be shown that $V_n$ is also given by $$ V_n = 2\sum_{m=0}^{n-1} \binom{n}{m+1} \binom{n-1}{m} v_n $$

MartinG
  • 978